High resolution images are desirable for a variety of applications such as in the areas of medical imaging, remote sensing, forensic imaging, and robotics applications, among others. Typically, high resolution images are obtained by combining information from a set of low resolution images obtained from two-dimensional scenes from different viewpoints. It also involves estimation of parametric relative motion. However, this technique may not be valid for three-dimensional scenes, in which case the motion estimation requires stereo disparity estimation.
Traditional stereo algorithms estimate disparity at the same resolution as that of the low resolution observed images which are assumed as averaged versions of a corresponding high resolution image. However, pixel averaging at low resolution may cause distortions in the estimated disparity. As a result, shape details may not be preserved in the high resolution depth map/stereo disparity map of the obtained image.